The invention relates to a novel
font special effect rendering method. The method comprises the following steps of randomly initializing a generated image, that is, randomly selecting a
white noise map of pixels and using the
white noise map as input of VGG-19; extracting high-dimensional content feature points of a
font image in a conv4_2 layer of VGG-19, and calculating the strongest gravitational content loss of a generated image and a
font image; extracting high-dimensional style feature points of style images of conv3_2 and conve_4
layers of the VGG-19 and calculating the strongest gravitational content loss of the generated image and the font image; solving a function of strongest gravitational content loss; minimizing value of a total
loss function; concluding iteration if the valueconverages or is less than an appointed threshold value; skipping to the step 2 otherwises. The method can reduce the possibility of incomplete rendering styles and wrongful rendering areas during arendering process.